What Merits Merit Pay?

Several of the replies to my previous posting discussed the power of the Teacher's Union over education, so I wanted to bring up one issue that is currently causing a contentious debate in California, the issue of Merit Pay for Teachers. I assume that this is one of the issues for which the anti-union folks criticize the Teachers Unions, who tend to be steadfast in their opposition to merit pay.

I believe that merit pay as a general concept sounds good, but I am not sure how it could actually work. I do not believe that it should be based on standardized test scores. As the unions love to point out (and I think that it is true), basing merit pay on test scores will punish teachers for teaching in the toughest schools. Even if it is based on improvement and not overall scores, certain groups are (by my guess) less likely to produce large amounts of progress. I'm not going to break it down by race or income because that is less conclusive, but I think we can all agree that the following groups will probably get lower show and less improvement: students who are frequently absent/truant, students with uneducated parents, learning disabled students, and recent immigrants who are still learning English. If we want to attract good teachers to work with these groups of students, I do not believe we should base merit pay on test scores. Another argument against this is that a teacher might make great progress with students who are far behind grade level and it still will not show up on the test. A student, for example, who comes into eighth grade at a third grade reading level and moves up to a sixth grade level by the end of the year has made great progress, but he will probably still bomb the eighth grade test.

The pro of standardized test scores is that they are quantifiable, objective, and easily ranked. If merit pay is not based on test scores, what else can they base it on? I think that it is pretty obvious which teachers are good and which teachers are not after spending only a short time in a teacher's classroom, but how to quantify and categorize that is a more difficult question.

Here are some ideas about what merit pay could be based on, each with its own problems:
1) Student, parent, principal evaluation- Highly subjective, a personal issue with a staff member could bias the process (in the principal's case). Students sometimes like teachers who let them slide by without learning a lot.
2) External or internal observation using Praxis III criteria- It probably takes a lot of resources to observe every teacher, and it should be done several times to really get a good idea of the teaching.
3) Teacher portfolios (teachers collect evidence and "artifacts" to prove that they are good teachers)- Once again, time consuming and somebody needs to review it, but a possiblity
4) Actual hours spent doing extra stuff (attending professional seminars and conferences, tutoring students before and after school, mentoring new teachers, serving on committees, etc.)- Probably a good idea, but then is it really merit pay or just extra pay for extra hours?
5) Extra pay for teaching high-needs subjects and/or in high-needs areas- A possibility, but how are schools supposed to prove their "high expectations" if they start offering teachers lots of extra money to attend? Doesn't that just prove that they don't believe in their school? And how do they prove which schools need these extra-high salaries? It might not actually be that hard to figure out these things if they just let the market demonstrate the areas of need. (The Middle School where I did my internship, for ex., has a bad reputation- largely unearned- in some circles, so several days they did not have enough substitute teachers.) On the other hand, would that make a poorly managed school eligible for higher salaries because nobody wants to work there?
6) Creating who new categories of teachers with higher salaries- Because a large factor in academic success is school attendance, I've heard the idea that they should create a category of "social worker teacher" who would be assigned to several low-attending students and would visit their homes, meet with parents, etc. to ensure that they attend school and take it seriously. This would be more work and might merit a higher salary.
6) Some combination- Perhaps there is some magical formula, but how much time would it take to calculate it out for each teacher?

For the record, this blog is not an endorsement or criticism of merit pay, I'm just trying to figure out how it would work best if it were implemented. Any ideas?